Analysis of Hot Topics in Cloud Computing

Analysis of Hot Topics in Cloud ComputingSpecial TopicU Deyi, Chen Guisheng, Zhang Haisu,Analysis of Hot Topics in CloudComputingLiDeyi', Chen Guisheng', Zhang Haisu2(1. Istite of Electronic System Engineering, Beiing 100141, P. A. China;2. PLA University of Science and Technology. Nanjing 210007, P. R. China)Abstract: In the field of cloud computing, topics such as computing resource virtualization, diftrences between grid and cloudcomputing, relationship between high-performance computers and cloud computing centers, and cloud security and standards haveattracted much research interest. This paper analyzes these topics and highlights that resource virualiation alows information servicesto be scalable, intensive, and specialized; grid computing involves using many computers for large-scale computing tasks, while cloudcomputing uses one platorm for multiple services; high-perormance computers may not be suitable for a cloud computing; security incloud computing focuses on trust management between service suppliers and users; and based on the existing standards,standardization of cloud computing should focus on iteroperability between services.Key Words: cloud computing; virualizatin; grid computing; security of cloud computing; standards of cloud computingn cloud computing, resources suchsuch as the technical basis, servicegave birth to cloud computing. Theas processing and storage aremodels, and commercial operation ofstructure and implementation details ofoftered for public use. Cloudcloud computing have already beena Web- based email administrationcomputing adopts a virtualized,widely discussedi”。system are virtualized when massdynamically scalable paradigm forThis article analyzes views onemails are sent, received, andorganizing, distributing, and usingresource virtualization, diferencesmanaged through browsers. Searchescomputing resources. In contrast to thebetween grid computing and cloudand matches are virtualized when atraditional paradigm, which relies oncomputing, relationship betweensearch engine responds to adesktop resources, computinghigh-performance computers andpersonalized search request. Networkresources provided in clouds are partcloud computing centers, and securityalbums are used for storing and sharingof social infrastructure and may have aand standards of cloud computing.pictures, and dynamic administration ofprofound impact on informationthe storage center is virtualized. Onlinetechnologies and their applications.1 Virtualization oftransaction and payment systems areThe development of cloud computing isalso virtualized.changing software engineering,Computing ResourcesComputing resources as virtualizedconfiguration of resources in networkWikipedia defines virtualization as theobjects can be classified according tocore and terminals, and acqition of*abstraction of computing resourcesM.computing capability, repositoryinformation and knowledge".Virtualization technologies emergedcapability, and interaction capability.Since the concept was proposed inearly in the history of technology. AnThese classifications correspond2007, cloud computing has beenoperating system, for example,exactly to CPU, repository, and input/promoted by academia and industryweakens the dependence of softwareoutput of a traditional computer.and has been transitioning from theoryapplication environments on hardwareBecause of resource virtualization,into practice. However, thepltorms, and even completely isolatescomputing is no longer treated as thedevelopment of cloud computingthe two. Middleware also weakens themain sector. If computing is dominant,technologies and their widespreaddependence of application s中国煤化工. s the computingimplementation will be a long-termsoftware operatin. Both areE _ory is dominant, theprocess because of the profoundof virtualization technology.HC N M H Ge repository center.impact that cloud computing serviceschange from stand-alone computers ToInteracions can also be treated as thewill bring to the public. Important topicsthe Internet, virtualization technologiesmain sector, while computing and( December 2010 val 8 No.4 z COMMUNICATIONS 01( Speciat TopicAnalysis of Hot Topics in Cloud Computing"_Li Deyi, Chen Guisheng, Zhang Haisurepository become supporting parts.Virtualization services on the Internetcan accommodate mass numbers ofusers to make highly personalized| Sener Clusters and Vtuolzciondemands through natural interaction.Vintual zed SenicesUsers need not consider the specificservice model of application softwareSever Tuseshpe Marketnor whether the service is beingsimultaneously leased by others. The4operating system of the computingplatform does not enter intoSever Iustesconsideration, nor does the physicalDato Sener SeulyServerconfiguration and administration ofEmalSeverbottom-layer resources such asserchSsever VdeoSenersoftware environment. The geographicBlowout of SeversSener Drvrsitaton Coused byposition of the computing center is also回Dedicaled Servicesirelevant to users. Virtualizationservices such as Software as a Service| Severs Underoking More Tasks(SaaS), Platform as a Service (PaaS)，Browso/soverand Infrastructure as a Service (laaS)Sorver Debutprovide dynamic and scalableClen/senerorganization, distribution, and utilizationof computing resources.Looking back at the transition ofresource configuration on the Internet,▲Figure 1. Transition of resource configuration on the Intermnet.as shown in Figure 1, can help inunderstanding the development ofproduction to intensive, scalable,multiple computers are networked intovirtualization services. The earliestspecialized production during thea grid to offer specific large -scaleserver appeared using the Client/ServerIndustrial Revolution. Today, thecomputing services- 'multiple for one."(C/S) structure- which was fllwed byinformation industry is also movingIn cloud computing, intensive andthe Browser/Server (B/S) structure.towards being intensive, scalable andspecialized Internet platforms are usedClient ends became thinner, andspecialized. Virtualization of computingto offer scalable services- "one forapplications could be directlyresources allows reasonablemultiple."accessed through the abstract of B/Sconfiguration and use of resources.lan Foster, the father of gridbrowser. Servers then emerged thatThe actual uilization of a single servercomputing, defined it as 'a hardwarecould perform diverse tasks, and manyis only around 15% whereas itilizationand software infrastructure thatinstitutions built servers that could hostof a server cluster can reach to moreprovides dependable, consistent,dedicated applications such as email,than 80%. This has direct implicationspervasive, and inexpensive access todata, security, and video. This led to afor energy saving, carbon emissionhigh- end computational capabiltie.!.o'"blowout in diversified servers. Such areduction, and green computing.Relying on private networks or thelarge number of servers prompted theInternet, grid computing organizes idlecreation of server trustees for alleviating2 Differences betweencomputer resources scattered inthe cost burden to owners ofdifferent areas and with diferentCloud Computing andmaintaining servers. However,capabilities and forms them into atrusteeship was not an effective solutionGrid Computingvirtual super computer by unifiedto intensive use of servers. If vitualizedBefore the emergence of cloudscheduling. In this way, complexservices could be offered and serverscomputing, grid computing'ol had beentasks- such as scientific taskschanged into "services,' the variousstudied for more than 10 years and hadrequiring mass data handling andkinds of computing resources could beattracted widespread attention. In theprocessing- can be completed. Thebetter integrated.first two years of cloud computing,basic application scenario of gridFrom the evolution of virtualizedgeneral opinion held that cloudcomputing involves integratingservices it can be seen that a servicecomputing was hot in the industrial fieldcomputing resources belonging toapproach to ofering computingbut cold in the academic fiell中国煤化工in different locations inresources was inevitable for cloudcomputing, the boot was onnger computingcomputing. This transition is similar infoot. What is the main differeMHCNMHGUIUuu culpuing tends to changenature to that of traditionalbetween cloud computing anu grtamanufacturing industry from masscomputing? Generally, in a grid systempowerful computing resources of02元COMUNICATIONS December 2010 val 8No4 ).Analysis of Hot Topics in Cloud ComputingSpecial TopicLi Deyi, Chen Guisheng, Zhang Haisu,certain Internet nodes into dynamic and(5) Cloud computing is designed totasks even enable computers at thescalable virtual resources. These aremeet the demands of mass numbers ofcloud computing center to be fixed tothen provided as services to massusers. People interact andtall metal racks with Velcro, making itnumbers of users. Cloud computingcommunicate with one another and areeasy to swap them out should theyprovides user-driven, on-demand andinvolved in computing processes.failz:." Through cooperation betweenpay-per-use services that may beSemantic handling and uncertaintyclusters, a search task involving bilinsdissolved. Its basic applicationprocessing are therefore required. Gridof microprocessor events and thescenario is oriented to the Internet, andcomputing is oriented to scientificreading of hundreds of megabytes ofa centralized computing resource poolcomputing; program input/output isdata can still be completed in severalis used to meet the mass and scatteredwell- defined according totenths of a second.demands of terminal users.specifications and instructions, andHigh-performance computers areThe biggest similarities between gridpeople are not generally part ofmainly used in science. The world's tenand cloud computing are resourcecomputing processes.most powerful computers in 2009 weresharing and virtual computing. Both ofIn short, cloud computing and gridmostly deployed in science researchthese provide users with sharedcomputing are applied in differentinstitutes, universities, and nationalInternet resources through virtualizationscenarios and have different goals.institutions. These include the Oakso that resources are reasonablyRidge National Laboratory, California;tilized.3 Performance of a Cloudthe Argonne National Laboratory of theHowever, they are different in someUS Depatment of Energy (DOE); therespects:Computing CenterUS National Institute for Computational(1) Cloud computing is based onA cloud computing center is basedSciences; Forschungs Zentrum Juelichcluster computing; nodes in cluster areon cluster computing. The large(FZJ), Germany; and the National Superself-governing and oriented to differentnumber of nodes at a cloud computingComputer Center in Tianin, China.service objects. Grid computing iscenter are interoperable and formThese high-performance computersbased on parallel computing;user-oriented virtual servers. Manyare applied in energy, manufacturing,cross-area computers are networkedinstitutions have purchasedweather forecast, nuclear,and a unified scheduling system ishigh-performance computers andhydromechanics, and astronomylts. Theused to distribute tasks to differentestablished computing centers.XT5 (Jaguar) in Oak Ridge is currentlynodes for handling.However, can these high-performancethe most powerful computer in the(2) Cloud computing recognizescomputers be used for cloudworld with a performance score of 1.75heterogeneity; that is, heterogeneity incomputing centers? Is a cloudPFlops in the Linpack test. It has nearlyprinciple, scale, and capability ofcomputing center the same as a250,000 computing cores, with anodes. Inter- -node resource sharing ishigh-performance computing center?theoretical peak computing rate of up toachieved through the interoperability ofWhat is the relationship between2.3 PFlops. An important goal ofthe service. Grid computing useshigh-performance computers is tomiddleware to shield heterogeneousvirtual servers of a cloud computingimprove computing speed in order tosystem for the upper layer applications,center?obtain higher performance parametersorienting users to the same environmentIf the largest, most intensive andin Linpack tests.for resource sharing.specialized cloud computing centersServices provided by a cloud(3) Cloud computing is oriented(owned by Google, Amazon, andcomputing center are oriented towardstowards persistent and diverseSalesforce) are considered, none of thethe diverse applications of large-scaleservices, and many cloud computingworld's ten most powerful computerssearch, network repository, andcenters in the Internet usually offerhave been used in the formation of theirnetwork business of mass numbers ofmany diverse services for differentservice clusters. It is thought thatusers. Therefore, cloud computingapplication fields. Grid computing isGoogle' s computing center consists ofcenters should be capable of providingprimarily used for completing one-off,more than 450,000 ordinary computershigh-quality service environments forspecific tasks that are pre-set.located in at least 25 places, while thetens of mllions of applications and(4) Cloud computing is operatedcomputing centers of Amazon andeffectively adapt to user demands andcommercially. providing best- -ffortSalesforce may be running on clusterservice innovation. Compared tomutilease services that are rented ftorsystems of 100,000 and 1000 ordinarysupercomputing centers. cloudagreed terms or charged onacomputers respectivly'". Since cloudcomputing centers do not follow thepay-per-use basis. Grid computingcomputing aims to meet independenttraditional task-oriented singlerelies on coordinated operationdemands of mass numbers C中国煤化工-1 but providebetween organizations. Bandwidth anddependence and intercrossirscalable, andperformance are guaranteed, however,between server cluster tasksMHC N M H Gces. Therefore,there are no obvious commercialresponding to different requesis Isilyl-uerurrnance computersmodels.greatly reduced. Such loosely coupleddeployed in high-pertormance( Decernber 2010 vol.8 No4 IzTE COMUNICATIONS 03( Special TopicAnalysis of Hot Topics in Cloud Computingli Deyi, Chen Guisheng, Zhang Haisucomputing centers are appropriate forcenters lacked the management,of www, Hypertext Transfer Protocolscientific problems requiring highmechanisms, and technical guarantee(HTTP) and Hypertext Markupconcurrency computing but are notof trust. Instead, data was saved inLanguage (HTML) enablednecessarily suitable for cloudprivate systems to prevent it beinginteroperation between terminals andcomputing.disclosed and to protect privacy. UsersWeb sites; Web browsers that usethemselves were responsible forthese protocols are able to seamlesslysystem security, often installing aaccess the www. Web services and4 Cloud Securityfirewall and antivirus and backing upService-Oriented Architecture (SOA)Resource sharing within clouddata. However, the rapid developmentopens the door for service computing.computing centers raises many securityof cloud computing will change thisResources in a cloud computingissues. Cloud computing is not a newpicture because users may not choosesystem exist in the form of services.weapon designed to solve securityto keep sensitive data themselves. TheMany commercial enterprises have builtissues. It is a computing paradigmcore model of cloud computing istheir platforms for cloud computing,based on the Internet and has some ofservice based on the premise that thereoffering data and services. However,the same security issues that exist inis a trust relationship between usersthe grammar and semantic differencescurrent information systems. Theseand service providers. The most basicbetween data and services still blockissues include viruses, maliciousand important guarantee foreffective information sharing andattacks, and information disclosureestablishing this relationship is theexchanges between them. Cloudwhen cloud services are being offered.bottom-up force created by thecomputing will not subvert existingTherefore, current information securitydemocratic nature of the Internet, whichstandards such as Simple Objecttechnologies may also be used tocannot be formed by a single personAccess Protocol (SOAP); Web Servicessecure cloud computing centers whilebut by interactions in communities.Description Language (WSDL); andadditional security measures areThus, trust is futilleldl not by one -offUniversal Description, Discovery, anddeveloped for cloud computing.testing or a set of fixed indicators but byIntegration (UDDI). However, CloudScalable, intensive, and specializedeliminating untrusted elements duringcomputing emphasizes serviceservices change the situation wherebythese interactions. This is a quality ofinteroperation based on existinginformation resources are scattered atcloud computing accumulated duringstandards. Designing higher levelend equipment. The development ofoperation. One of the key issues to beprotocols and specifications forSecurity as a Service (SECaaS) issolved in trust administration of cloudopenness and interoperation isexpected to improve Internet security.systems is how to best abstract andextremely important in order to aciltateIntensive and specialized securityapply the trust that emerges during theinteroperation between clouds andservices may be implemented in cloudevolution of a cloud computing system.interoperation between services andcomputing centers, possibly eliminatingSociety, poltics, and technology canend users.the need for end users to installjointly promote the establishment,The International Standardizationpatches and kill viruses. Backup maymaintenance, and administration ofOrganization ISO/IEC JTC1 SC32 isbe treated as a kind of service fortrust in cloud computing.responsible for the development ofimplementing specific cloud backupISO/IEC 19763-Metamodelservices.5 Standardization of CloudFramework for Interoperability (MFI).Therefore, the security focus in cloudMFI is a standard family that providesservices will gradually turn to trustComputingreference for the model registration,management. Traditional informationCloud computing provides users withontology registration, and modelsecurity wil develop into trustmany types of services with variablemapping of registered information datamanagement between servicegranularity. Interconnection,resources. It promotes interoperation ofproviders and users. The analogy of ainterworking, and interoperationsoftware services. China participated inbank deposit can be used to describebetween these services is a keythe development of ISO/IEC 19763-3 inthe trust relationship between userstechnical foundation for implementingthe standard famiy. ISO/EC JTC1 SC7and cloud service centers. In the past,an open cloud platform. In fact,and ISO/IEC JTC1 SC38 establishedmany people thought it was safest tointerconnection, interworking, andcloud computing study groups in 2009,hide their money away in secret places.interoperation are the basicwith the tasks of developingHowever, with the development ofcharacteristics of networks in any stageterminology related to cloud computingbanks, people signed service contractsof development. Protocols for Localand drafting study reports on cloudto have their wealth safely stored andArea Networks (LAN) and Wide Areacomputing standardization.managed. Sensitive information isNetworks (WAN) are used f中国煤化工- nizations such as thesimilar to wealth. In the signal computerinterworking of computing CAlince (CSA)", Openage and the early days of Internet,Transport Control ProtocolYHCNMHGmi",andCloudleaving sensitive data in networkProtocol (TCP/IP) enablesComputng Interoperability Forumservice centers was risky because theinter-network connection. In the days(CIF)*HI are taking steps to develop04zTE COMMUNICATIONS December 2010 Vol.8 No.4Analysis of Hot Topics in Cloud ComputingSpecial TopicLi Deyi, Chen Guisheng, Zhang HaisuPrelevant cloud computing standardsUSA: EEE 2008 10p.Biographies[4] GERMAIN-RENAUD C. RANAOF. Thefor virtual machine image distibution,Convergence of Clouds. Grids, and Autonomics [J].Li Deyivirtual machine deployment andISIERICKSONJS SPENCE S RHODESM e acontrol, communication between virtualFCortatCetereCollaboratioo Sacesin h。LiDeyi is an academic at themachines within a cloud, persistentCloud [UJ IEEE Internet Computing 2009, 13(5):Chinese Academy ofEngineering (CAE). andarepository, and safe vitual machine34-42.professor at the Instilute of6]LEIBAB. Having One's Head in the Cloud [J]. IEEEconfiguration. These organizations areIntermet Computing. 2009. 13(5): 4-6.ElecroncSystem Heahead of international standardization(7IJENSEN M. SCHWENK J. GRUSCHKAN.etal Onresearch interests includeTechnical Security Issues in Cloud Computing (CWorganizations in the development ofcompulter engineering.cloud computing standerds. The Chinaartificial itelligence, and21-25. 200 Bangalre. Ioua. LoAanan..pcommand automation.Cloud Computing Technology andUSA: IEEE Computer Society. 2009: 109-116Industry Aliance (CCCTIA) also aims to[8)Cloud Computing (EB/OL] (2009-05- 23]. ht:/n..contribute to the standardization ofwkipedia. org/wikuCloud. .compuling.Chen Guisheng[9]祁金华.国内虚拟化案例创析[NI.网络世界，cloud computing.2007-07-02.Ql Jnhua. Analysis of Domestic Vrtualization CasesChen Guisheng is a senior(10] FOSTER i. KESSELMAN C. The Gnd: Blueprint forengineer at the Institute of .“a NewComo tino IfrstuctureIMl. SanFrancisco, CA. USA: Morgan KautmannEngineering. China. HisPublishers, 1999[11] The Etcent Cloud: Al of Salestorce Runs on Onlyarificial itelligence, and1000 servers [EB/OL]. [2009-05-23] ht://techcrunch com/200/03/23/the- eficient loudall-ol- salesforce-runs-on-oReferences[12)尼古拉斯.卡尔IT不再重要:互联网大转换的制高点[1]李德毅张海粟.腿出图灵机的云计算[J.中国计算机一云计页(MI.闫鲜宁，泽北京中出出版社。2008.Zhang Haisu学会通讯。 200(12)NICHOLAS Carr. The Big Switch; Rewiring theu Deyi. ZHANG Haisu. Cloud Computation BeyondWorld, from Edison to Google [M]. TranslatedbyTuring Machines [UI Communicatons of CCF.2009Yan Xiannng. Beyng: CITIC Press, 200813] TOP500 Supercomputing Sites (EB/OL].Zhang Haisu is a teaching[2] ARMBRUST M. FOX A. GRIFFITH A. et al. Above the[2009-06- 25] http /www top500 org.assistant al the PLAClouds. A Berkeley View of Cloud Computing [R].14] Cloud Security Aliance (CSA)- -Security BestUniversity of Science andBerkeley. CA, USA. Distrbuted Systems Lab,Practces for Cloud.org[EB/OL (2009 -06-28].[3] FOSTER i, ZHAO Yong. RAICUI, et al. CloudI 15] Open Cloud Consorium (EB/OL]. [200 _08- -25] .CompiexnetMorks ankoComputing and Gnd Compuing 360-Degreehtp:/:pe/clococonsortourorng/.computer engineering.Compared [C)/Proceedings of the IEE Grid[16] Cloud Compuling Iteroperability Forum (CCIF)Computing Environments Workshop (GCE 08). Nov[EB/OL] [2009-06-25] h/::/w.wooruo.12- 16.20080 Austin, TX. USA. Piscataway. NJ,org/.RoundupNew Member Biography of ZTE Communications Editorial BoardCheng-Zhong Xu received his BSand MS degrees from Nanjingprojects in these areas. His research was also supported by otherUniversity in 1986 and 1989,funding agencies like U.S. NASA and NSF of China. He hasrespectively, and Ph.D. degree frompublished more than 160 scientific papers, including over 20 articlesthe University of Hong Kong in 1993, alin EEE/ACM transactions. He is the author of "Scalable and Securein computer science. He is currently aInternet Services and Architecture* (Chapman & Hal/CRC Press,professor in the Department of2005) and a co-author of "Load Balancing in Parallel Computers:Electrical and Computer Engineering of Theory and Practice" (Kluwer Academic, 1996).Wayne State University, the Director of Dr. Xu serves on the editorial boards of IEEE Transactions onCloud and Internet ComputingParallel and Distributed Systems, Journal of Parallel and DistributedLaboratory, and the Director of SunComputing. Journal of Parallel, Emergent, and Distributed Systems,Microsystems' Center of Excellence inJournal of Computers and Applications, and Journal of HighOpen Source Computing andPerformance Computing and Networking. He has served dozens ofApplications. He also holds a visitinginternational conferences and workshops in capacities as programprofessorship at Shanghai Jiaotongchair, general chair, and plenary speaker. He is a recipient of the中国煤化工ity in 2000, theHis research interest is mainly in distributed systems and the Internet, PresidenH in 2002, and thedatacenters and cloud computing, and wireless embedded systems, CareerDYHi is also a recipientofwith an emphasis on resource and system management forthe 'Over.CNMHGisFoChina.Formoreperformance, avlabiliy, rlibili, energy eficincy, and security.information, please visi: htp:/://w ece .eng.wayne .edu/~Czxu/.( December 2010 Vol.8 No.4 I[ ZπTE COMMUNICATIONS05